Get important education news and analysis delivered straight to your inbox
For much of the previous decade, advocates of education technology imagined a classroom where computer algorithms would differentiate instruction for each student, delivering just the right lessons at the right time, like a personal tutor. The evidence that students learn better this way has not been strong and, instead, we’re reading reports that technology use at school sometimes hurts student achievement.
So it was interesting to see McKinsey & Co., an elite consulting firm, reframe the argument for buying education technology away from computerized instruction to something more pedestrian: saving teachers time. A January 2020 report by the firm estimated that between 20 and 40 percent of the 50 hours that a typical teacher currently works a week could be saved through existing automation technology, often enabled by artificial intelligence (AI). That adds up to 13 saved hours a week, hours of freedom that could help relieve teacher burnout. Those hours could also be reallocated so that teachers can do more of what teachers do best: interact with students.
“Many of the attributes that make good teachers great are the very things that AI or other technology fails to emulate: inspiring students, building positive school and class climates, resolving conflicts, creating connection and belonging, seeing the world from the perspective of individual students, and mentoring and coaching students,” the McKinsey authors wrote. “These things represent the heart of a teacher’s work and cannot — and should not — be automated.”
The McKinsey authors suggest that existing technology can be used to help teachers in several areas: planning lessons, assessing students, grading homework, giving feedback and administrative paperwork. The consultants aren’t suggesting that computers can replace any of these tasks entirely but rather reduce the amount of time teachers have to spend on them. For example, they estimate that lesson preparation could be cut from almost 11 to six hours. They calculate that weekly grading could be cut in half from six to three hours. And they say that two hours a week of administrative paperwork could be trimmed.
The report is provocatively titled, “How artificial intelligence will impact K-12 teachers,” though many of the recommendations are for categories of software applications that don’t necessarily use sophisticated AI algorithms at all, such as sites where teachers can find curriculum materials posted by other teachers. I was curious to learn what scientists who are involved in studying and developing AI in education thought of McKinsey’s analysis and heard a range of praise, skepticism and outright criticism.
A lot of automatic grading technology isn’t very good yet, AI experts told me. “There’s stuff out there than you can use tomorrow but I also think there’s still a lot to be done,” said Ryan Baker, a professor at the University of Pennsylvania who studies how students learn from educational software. Computers can easily grade math computations, he said, but automated writing feedback or feedback for more complicated math or science problems still needs to get better.
Another shortcoming with a lot of existing technology, Baker said, is using the data that computerized systems generate for lesson planning. If computers are grading homework or assessing what students know, the systems need to convey results for each student in a way that’s useful for teachers. “It’s a hard challenge,” he said. “There’s a lot to be done in taking sophisticated AI models that I work with and translate them into something that’s understandable to teachers and that they trust.”
Often developers of educational software create dashboards for teachers to decipher that sometimes add to their workloads instead of saving them time. Or the feedback for teachers is very simplistic, such as, 56 percent of the class got a particular homework problem wrong. But there’s no insight into how to help each student.
Stefania Druga, a doctoral student at the University of Washington, who studies AI in education and who founded a coding project to teach students how AI models work, argues that automated grading can have the unintended consequence of breaking the feedback loop between teacher and student. Often a student can trick a robograder and get the problem right without understanding the underlying concept, she explained. Or the student learns what metrics the automatic grader looks at and “optimizes” for them. For example, a writing feedback program might emphasize the use of the word “evidence” and other synonyms and give high marks to incomprehensible essays that sprinkle those keywords throughout. Though time consuming, having teachers actually read student work directly, she says, is important for the learning process.
Some of the biggest time savings, according to McKinsey, could be in using existing technology for lesson planning. But curriculum experts who reviewed popular online curriculum resources found them to be mediocre. McKinsey acknowledges that quality in ed tech is a problem. Schools and teachers are pitched a “myriad of competing solutions,” some of which “promise great things but deliver little,” the report explains. McKinsey called for “neutral arbiters” to evaluate the quality of software with “objective and rigorous performance data.”
The cynic in me was wondering if McKinsey is merely offering ed tech companies new talking points to sell their existing wares. Instead of boasting about how much they boost student performance, ed tech marketers can tout how many teacher hours they save. But I was intrigued with McKinsey’s suggestion to use AI more for the back office. The consultants suggest that automation software could fill out forms or suggest potential responses, maintain inventories of materials, equipment, and products, and automatically order replacements.
Andrew Berning, the president of the Renaissance Institute, a company that sells back-office tech services to schools, told me that McKinsey’s reframing is “spot on.” “The market has been too focused on developing AI ‘teaching machines’ and not enough in the area where we can really have an impact, such as facilitation, automation and teachers support,” he said, via e-mail. His big growth area: tools for schools to monitor their internet traffic to detect cyber threats, such as phishing, fraud and ransomware.
That’s what schools need: more technology to protect them from the harm that the technology they’ve already bought is causing.
This story about AI in education was written by Jill Barshay and produced by The Hechinger Report, a nonprofit, independent news organization focused on inequality and innovation in education. Sign up for the Hechinger newsletter.